Re: [ft] Extracting sub-glyph transformation

[Kevin Rogovin]

Hi,

Thankyuo for the fast reply. The part that I am having trickiness is when
flags has the bit ARGS_ARE_XY_VALUES as down; then the values of args1 and
arg2 are indices into an array, the relevant code I found is
src/truetype/ttgload.c, function TT_Process_Composite_Component():

[code]
      FT_UInt     num_points = (FT_UInt)gloader->base.outline.n_points;
      FT_UInt     k = (FT_UInt)subglyph->arg1;
      FT_UInt     l = (FT_UInt)subglyph->arg2;
      FT_Vector*  p1;
      FT_Vector*  p2;


      /* match l-th point of the newly loaded component to the k-th point */
      /* of the previously loaded components.                             */

      /* change to the point numbers used by our outline */
      k += start_point;
      l += num_base_points;
      if ( k >= num_base_points ||
           l >= num_points      )
        return FT_THROW( Invalid_Composite );

      p1 = gloader->base.outline.points + k;
      p2 = gloader->base.outline.points + l;

      x = p1->x - p2->x;
      y = p1->y - p2->y;
[/code]

I see that  arg1 and arg2 function as indices when the bit
ARGS_ARE_XY_VALUES are down, but I don't know how to get those vertices
from the index values. And that is likely the easy part... for then I need
to copy the tricky code to adjust the translations too.

Would it be possible to add a new entry point to FreeType that returns the
computed translation? Users will still likely want the flags (for example
to have a fast check if the FT_Matrix is identity, scaling only or 2x2).
Related question: are the values of the FT_Matrix returned by
FT_GetSubGlyphInfo() in units of the parent glyph?

Best Regards,
 -Kevin Rogovin

On Fri, Aug 31, 2018 at 8:00 AM Werner LEMBERG <address@hidden> wrote:

>
> > I am trying to load glyph’s geometric data with FT_LOAD_NO_RECURSE.
> > The entry point FT_GetSubGlyphInfo() has almost what I needed, but
> > not quite.  It does provide the 2x2 matrix, but not the translation
> > (at least when I read the docs and FreeType source code).  [...]
>
> The FreeType documentation explicitly refers to the OpenType
> specification; for clarity, I've now added a link.
>
> I hope this helps.  It's quite tricky, indeed.
>
>
>    Werner
>